Ngày đăng: 05/11/2019, 15:54
32 Editors SMART INNOVATION, SYSTEMS AND TECHNOLOGIES Lakhmi C Jain Himansu Sekhar Behera Jyotsna Kumar Mandal Durga Prasad Mohapatra Computational Intelligence in Data Mining - Volume Proceedings of the International Conference on CIDM, 20-21 December 2014 Smart Innovation, Systems and Technologies Volume 32 Series editors Robert J Howlett, KES International, Shoreham-by-Sea, UK e-mail: rjhowlett@kesinternational.org Lakhmi C Jain, University of Canberra, Canberra, Australia, and University of South Australia, Adelaide, Australia e-mail: Lakhmi.jain@unisa.edu.au About this Series The Smart Innovation, Systems and Technologies book series encompasses the topics of knowledge, intelligence, innovation and sustainability The aim of the series is to make available a platform for the publication of books on all aspects of single and multi-disciplinary research on these themes in order to make the latest results available in a readily-accessible form Volumes on interdisciplinary research combining two or more of these areas is particularly sought The series covers systems and paradigms that employ knowledge and intelligence in a broad sense Its scope is systems having embedded knowledge and intelligence, which may be applied to the solution of world problems in industry, the environment and the community It also focusses on the knowledgetransfer methodologies and innovation strategies employed to make this happen effectively The combination of intelligent systems tools and a broad range of applications introduces a need for a synergy of disciplines from science, technology, business and the humanities The series will include conference proceedings, edited collections, monographs, handbooks, reference books, and other relevant types of book in areas of science and technology where smart systems and technologies can offer innovative solutions High quality content is an essential feature for all book proposals accepted for the series It is expected that editors of all accepted volumes will ensure that contributions are subjected to an appropriate level of reviewing process and adhere to KES quality principles More information about this series at http://www.springer.com/series/8767 Lakhmi C Jain Himansu Sekhar Behera Jyotsna Kumar Mandal Durga Prasad Mohapatra • • Editors Computational Intelligence in Data Mining - Volume Proceedings of the International Conference on CIDM, 20-21 December 2014 123 Editors Lakhmi C Jain University of Canberra Canberra Australia and University of South Australia Adelaide, SA Australia Himansu Sekhar Behera Department of Computer Science and Engineering Veer Surendra Sai University of Technology Sambalpur, Odisha India Jyotsna Kumar Mandal Department of Computer Science and Engineering Kalyani University Nadia, West Bengal India Durga Prasad Mohapatra Department of Computer Science and Engineering National Institute of Technology Rourkela Rourkela India ISSN 2190-3018 ISSN 2190-3026 (electronic) Smart Innovation, Systems and Technologies ISBN 978-81-322-2207-1 ISBN 978-81-322-2208-8 (eBook) DOI 10.1007/978-81-322-2208-8 Library of Congress Control Number: 2014956493 Springer New Delhi Heidelberg New York Dordrecht London © Springer India 2015 This work is subject to copyright All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed The use of general descriptive names, registered names, trademarks, service marks, etc in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made Printed on acid-free paper Springer (India) Pvt Ltd is part of Springer Science+Business Media (www.springer.com) Preface The First International Conference on “Computational Intelligence in Data Mining (ICCIDM-2014)” was hosted and organized jointly by the Department of Computer Science and Engineering, Information Technology and MCA, Veer Surendra Sai University of Technology, Burla, Sambalpur, Odisha, India between 20 and 21 December 2014 ICCIDM is an international interdisciplinary conference covering research and developments in the ﬁelds of Data Mining, Computational Intelligence, Soft Computing, Machine Learning, Fuzzy Logic, and a lot more More than 550 prospective authors had submitted their research papers to the conference ICCIDM selected 192 papers after a double blind peer review process by experienced subject expertise reviewers chosen from the country and abroad The proceedings of ICCIDM is a nice collection of interdisciplinary papers concerned in various proliﬁc research areas of Data Mining and Computational Intelligence It has been an honor for us to have the chance to edit the proceedings We have enjoyed considerably working in cooperation with the International Advisory, Program, and Technical Committees to call for papers, review papers, and ﬁnalize papers to be included in the proceedings This International Conference ICCIDM aims at encompassing a new breed of engineers, technologists making it a crest of global success It will also educate the youth to move ahead for inventing something that will lead to great success This year’s program includes an exciting collection of contributions resulting from a successful call for papers The selected papers have been divided into thematic areas including both review and research papers which highlight the current focus of Computational Intelligence Techniques in Data Mining The conference aims at creating a forum for further discussion for an integrated information ﬁeld incorporating a series of technical issues in the frontier analysis and design aspects of different alliances in the related ﬁeld of Intelligent computing and others Therefore the call for paper was on three major themes like Methods, Algorithms, and Models in Data mining and Machine learning, Advance Computing and Applications Further, papers discussing the issues and applications related to the theme of the conference were also welcomed at ICCIDM v vi Preface The proceedings of ICCIDM have been released to mark this great day in ICCIDM which is a collection of ideas and perspectives on different issues and some new thoughts on various ﬁelds of Intelligent Computing We hope the author’s own research and opinions add value to it First and foremost are the authors of papers, columns, and editorials whose works have made the conference a great success We had a great time putting together this proceedings The ICCIDM conference and proceedings are a credit to a large group of people and everyone should be there for the outcome We extend our deep sense of gratitude to all for their warm encouragement, inspiration, and continuous support for making it possible Hope all of us will appreciate the good contributions made and justify our efforts Acknowledgments The theme and relevance of ICCIDM has attracted more than 550 researchers/ academicians around the globe, which enabled us to select good quality papers and serve to demonstrate the popularity of the ICCIDM conference for sharing ideas and research ﬁndings with truly national and international communities Thanks to all who have contributed in producing such a comprehensive conference proceedings of ICCIDM The organizing committee believes and trusts that we have been true to the spirit of collegiality that members of ICCIDM value, even as maintaining an elevated standard as we have reviewed papers, provided feedback, and present a strong body of published work in this collection of proceedings Thanks to all the members of the Organizing committee for their heartfelt support and cooperation It has been an honor for us to edit the proceedings We have enjoyed considerably working in cooperation with the International Advisory, Program, and Technical Committees to call for papers, review papers, and ﬁnalize papers to be included in the proceedings We express our sincere thanks and obligations to the benign reviewers for sparing their valuable time and effort in reviewing the papers along with suggestions and appreciation in improvising the presentation, quality, and content of this proceedings Without this commitment it would not be possible to have the important reviewer status assigned to papers in the proceedings The eminence of these papers is an accolade to the authors and also to the reviewers who have guided for indispensable perfection We would like to gratefully acknowledge the enthusiastic guidance and continuous support of Prof (Dr.) Lakhmi Jain, as and when it was needed as well as adjudicating on those difﬁcult decisions in the preparation of the proceedings and impetus to our efforts to publish this proceeding Last but not the least, the editorial members of Springer Publishing deserve a special mention and our sincere thanks to them not only for making our dream come true in the shape of this proceedings, but also for its brilliant get-up and in-time publication in Smart, Innovation, System and Technologies, Springer vii viii Acknowledgments I feel honored to express my deep sense of gratitude to all members of International Advisory Committee, Technical Committee, Program Committee, Organizing Committee, and Editorial Committee members of ICCIDM for their unconditional support and cooperation The ICCIDM conference and proceedings are a credit to a large group of people and everyone should be proud of the outcome Himansu Sekhar Behera About the Conference The International Conference on “Computational Intelligence in Data Mining” (ICCIDM-2014) has been established itself as one of the leading and prestigious conference which will facilitate cross-cooperation across the diverse regional research communities within India as well as with other International regional research programs and partners Such an active dialogue and discussion among International and National research communities is required to address many new trends and challenges and applications of Computational Intelligence in the ﬁeld of Science, Engineering and Technology ICCIDM 2014 is endowed with an opportune forum and a vibrant platform for researchers, academicians, scientists, and practitioners to share their original research ﬁndings and practical development experiences on the new challenges and budding confronting issues The conference aims to: • Provide an insight into current strength and weaknesses of current applications as well as research ﬁndings of both Computational Intelligence and Data Mining • Improve the exchange of ideas and coherence between the various Computational Intelligence Methods • Enhance the relevance and exploitation of data mining application areas for enduser as well as novice user application • Bridge research with practice that will lead to a fruitful platform for the development of Computational Intelligence in Data mining for researchers and practitioners • Promote novel high quality research ﬁndings and innovative solutions to the challenging problems in Intelligent Computing • Make a tangible contribution to some innovative ﬁndings in the ﬁeld of data mining • Provide research recommendations for future assessment reports ix Accelerated FFT Computation for GNU Radio Using … 661 2.1 FFT in GPU of Raspberry Pi The use of processing power of GPU for calculating FFT can reduce the computational limitation of normal CPU Andrew Holme has designed such library which uses the GPU for calculating the FFT in Raspberry Pi [1] This library is purely open source and can be installed on to Raspberry Pi platform by running its command in command prompt of Raspberry Pi We implemented code for FFT computation of an input signal of given length on this library The code has three main functions GPU_FFT_prepare that creates structure for storing the input The memory is effectively utilized by using a maloc function, GPU_FFT_execute is the core function that is used to execute the FFT and also it computes the time of computation, and ﬁnally GPU_FFT_release that will release all the previously stored memory locations for further reuse The program is written in such a way that it take input of size as power of 2, and also takes the number of batches containing the number of arrays of input Each row of the batch corresponds to each input and result is calculated row wise This will return the FFT of each row along with the time it took to calculate the same Any signal that has been generated randomly is given as its input to compute its FFT This is compared with the previously generated time of execution obtained from Raspberry Pi CPU and MATLAB software 2.2 FFT in CPU of Raspberry Pi The CPU in Raspberry Pi belongs to 700 MHz ARM1176JZF-S core It is more similar in conﬁguration to that of old Pentium II processor, which is comparatively slower as compared with the newer versions available and hence it cannot provide much of execution speed The execution power in the CPU cannot be up to the expectation while computing a high mathematical computation incorporated FFTs FFT is ideally used to make DFT having computation of OðN Þ faster FFT has number of computation of OðNlog2 NÞ This is incorporated into CPU of Raspberry Pi using executables like FFT_processor, GNU plot_driver, Signal_Source and Sound_Source The time taken for execution of a real time signal is computed for various lengths of the signal and is tabulated for comparison The real time input signal is shown in Fig The obtained FFT plot is depicted in Fig 2.3 FFT in Intel-COREi5 FFT computed on intel-COREi5 is done using MATLAB software version R2013a (64 bit) The processor is having a clock speed of 3.3 GHz and maximum turbo frequency of 3.7 GHz MATLAB has got capability of reading any one dimensional signal which can be passed as the input for computing FFT We generated a signal 662 S Sabarinath et al Fig Noisy input signal in time domain Fig Frequency spectrum using this software and passed as the input of FFT function MATLAB software already has inbuilt FFT command which is used to calculate the FFT Time of computation for each input is calculated within MATLAB itself Batches of inputs are given each batch having a particular number of inputs So a batch will contain two inputs The computation time of both these inputs are calculated together in a batch Accelerated FFT Computation for GNU Radio Using … 663 Results and Discussion A comparison of time taken for FFT computation by CPU, GPU and MATLAB software is done here Signal having various lengths are considered Here we consider length 256, 512, 1,024 and 2,048 Each signal contains four batches, each batch having array of input signal whose FFT need to be computed Results obtained are tabulated for easy comparison Table shows the execution time for FFT in CPU of Raspberry Pi, Table shows the execution time for FFT in intelCOREi5 The computation of FFT using GNU radio is found to be time consuming, as it is done using GPU in Raspberry Pi This paper presents an effective solution to the above mentioned problem From detailed analysis of sequences having various lengths (for different batches), it is found that CPU of Raspberry Pi consumes much more time as compared with MATLAB software which runs in conventional CPU as shown in Tables and But at the same time GPU of Raspberry Pi performs faster FFT computation than both of the above methods Execution time of GPU is found to come in the order of micro seconds, which is shown in Table Table Execution time of FFT in CPU of Raspberry Pi given in seconds Batches Length of FFT 256 512 1,024 2,048 0.2235 0.4229 0.6602 0.9298 0.3134 0.7378 1.08 0.534 1.817 1.20 2.448 1.0335 2.1096 3.1788 4.185 Table Execution time of FFT in GPU of Raspberry Pi given in seconds Batches Length of FFT 256 512 1,024 2,048 0.23 × 10−6 46 × 10−6 72 × 10−6 114 × 10−6 36 × 10−6 76 × 10−6 113 × 10−6 154 × 10−6 57 × 10−6 112 × 10−6 174 × 10−6 238 × 10−6 117 228 346 455 Table Execution time of FFT in intel-COREi5 given in seconds × × × × 10−6 10−6 10−6 10−6 Batches Length of FFT 256 512 1,024 2,048 0.014089 0.039796 0.036306 0.098299 0.038113 0.095895 0.167499 0.149205 0.086689 0.188929 0.278827 0.320916 0.052547 0.043302 0.075456 0.081684 664 S Sabarinath et al Conclusion and Future Work In this paper, a comparison is done on the speed of FFT computation in GPU with that of CPU of Raspberry Pi and computation performed in intel-COREi5 processor (using MATLAB software) It was found that GPU allows us to compute the FFT much faster than both CPU of Raspberry Pi and intel-COREi5 processor In future this GPU can be synchronized with normal processor in such a way that whenever faster computational power is needed, processor automatically shifts to GPU The motivation behind doing this work was the slower computation speed of FFT in GNU Radio Companion The slow computational speed is compensated by the use of GPU in Raspberry Pi This computation can even be performed on environments where there is absence of computer, since Raspberry Pi itself act as a mini computer As a future work this FFT computation could be utilized in computer vision based navigation system using Raspberry Pi for path tracking applications References Duhamel, P., Vetterli, M.: Fast Fourier transforms: a tutorial review and a state of the art Sig Process 19, 259–299 (1990) Accelerating Fourier transforms using the GPU | Raspberry http://www.raspberrypi.org/ accelerating-fourier-transforms-using-the-gpu/ WhatIsGR—GNU Radio—gnuradio.org http://gnuradio.org/redmine/projects/gnuradio/wiki/ WhatIsGR/ Gandhiraj, R., Soman, K.P.: Modern analog and digital communication systems development using GNU radio with USRP Telecommun Syst 1-15 (2013) Brigham, E.O., Yuen, C.: The fast Fourier transform Syst Man Cybern IEEE Trans 8, 146–146 (1978) Sorensen, H.V., Jones, D.L., Heideman, M., Burrus, C.S.: Real-valued fast Fourier transform algorithms Acoust Speech Sig Process IEEE Trans 35, 849–863 (1987) Gandhiraj, R., Ram, R., Soman, K.P.: Analog and digital modulation toolkit for software deﬁned radio Proc Eng 30, 1155–1162 (2012) Feature Extraction and Performance Analysis of EEG Signal Using S-Transform Monorama Swain, Rutuparna Panda, Himansu Mahapatra and Sneha Tibrewal Abstract Feature can be described as a functional component observed from a data set The extracted features give the information related to a signal, thus it requires to calculate cost of information processing and complexity of analyzing a huge data set This paper presents a feature extraction method using S transform Five data sets are taken and feature extraction has been performed by implementing two methods: ﬁrst by applying S-transform and other without S-transform The performance of the neural model is evaluated on the basis of training performance and classiﬁcation accuracies and the results conﬁrmed that the proposed scheme has potential in classifying the EEG signals Keywords Feature abstraction EEG signal Á Neural network Á S-transform Introduction The German psychiatrist, Hans Berger performed the ﬁrst electroencephalographic (EEG) recording in humans (Berger 1929) EEG signals are electrical signals obtained due to brain activity These signals are then processed by a computer Feature extraction is required for construction of an efﬁcient brain computer interface EEG signal processing follows a three step procedure The ﬁrst step M Swain (&) Á H Mahapatra Á S Tibrewal Silicon Institute of Technology, Bhubaneswar, India e-mail: mswain@silicon.ac.in H Mahapatra e-mail: rintusilicon@gmail.com S Tibrewal e-mail: tibrewalsneha24@gmail.com R Panda Department of ECE, Veer Surendra Sai University of Technology, Sambalpur, India e-mail: r_ppanda@yahoo.co.in © Springer India 2015 L.C Jain et al (eds.), Computational Intelligence in Data Mining - Volume 2, Smart Innovation, Systems and Technologies 32, DOI 10.1007/978-81-322-2208-8_61 665 666 M Swain et al acquisition of brain signals and subsequent processing to remove the unwanted noise components Secondly a distinct feature is identiﬁed and extracted In the ﬁnal step machine learning algorithms are used for pattern recognition A number of signal processing techniques have been widely used to study various EEG signals after the ﬁrst EEG record [1, 2] was discovered In the past few years, researchers in applied mathematics and signal processing have developed many methods like wavelet transform [3, 4] and S transform for the extraction of features of brain signals Feature extraction techniques are required to distinguish the requisite signals from background Numerous signal processing techniques have been applied to obtain representations and extract the features of a signal Feature extraction simpliﬁes the analysis of complex data by reducing data dimensions Careful selection of features is an important criterion to obtain discriminative information for classiﬁcation Recent advances in the ﬁeld of neural networks have made them attractive for analyzing signals Artiﬁcial neural networks [5] provide an insight into the functioning of real network of neurons The application of neural networks has opened a new area of solving problems not resolvable by other signal processing techniques ANNSs not only model the signal, but also make a decision as to the classify the signal Many methods of feature extraction have been applied to extract the relevant characteristics from a given EEG data Here our objective is to analyze the EEG signals and classify the EEG data into different classes Our main target is to improve the accuracy of EEG signals Analysis of EEG signals provides a crucial tool for diagnosis of neurobiological diseases [6] The problem of EEG signal classiﬁcation [7] into healthy and pathological cases is primarily a pattern recognition problem using extracted features The rest of the paper is organized as follows: Sect describes the database and the techniques used in our work, Discrete S transform and artiﬁcial neural network, Sect gives the feature extraction and classiﬁcation results Data Acquisition and Methods 2.1 Data Selection and Recording We have used the publicly available data with Bonn University The complete dataset consists of ﬁve classes viz A, B, C, D, E each of which contains 100 single channel EEG segments of 23.6 s duration and having 4,097 data points Sets A and B consists of signals taken from surface EEG recordings that were carried out on ﬁve healthy volunteers using a standardized electrode placement scheme given in Fig Data in set A belongs to volunteers relaxed in an awakened state with eyes open while data in set B belongs to the same group of individuals having eyes closed Set C–E contains data from ﬁve patients all of whom have achieved complete seizure control Set C contains data from the hippocampal formation of the opposite hemisphere of the brain In Set D recordings were taken from the epileptic zone but Feature Extraction and Performance Analysis of EEG Signal … 667 Fig The 10–20 system of electrode placement during seizure free interval while Set E contained only seizure activity From the data available, a rectangular window of length 256 discrete data was selected to form a single EEG segment 2.2 Analysis Using S Transform The discrete S-transform is dened as follows: Let, qẵkT ; k ẳ 0; 1; ; L À 1, denote a discrete time series corresponding to a signal q(t) with a time sampling interval of T [8] The discrete Fourier transform of the signal can be obtained as follows: Q LÀ1 h n i 1X qẵkT ei2Pnk=Lị ẳ LT L kẳ0 1ị where n ¼ 0; 1; ; L À and the inverse discrete Fourier transform is q½kT ¼ LÀ1 h X n i i2Pnk=L Q e LT nẳ0 2ị In the discrete case, the S-Transform is the projection of the vector deﬁned by the time series q[kT], onto a spanning set of vectors The spanning vectors are not 668 M Swain et al orthogonal and the elements of the S-Transform [8] are not independent Each basis vector (of the Fourier transform) is divided into L localized vectors by an elementby-element product with the L shifted Gaussians, such that the sum of these L localized vectors is original basis vector The S-Transform of a discrete time series q[kT], is given by S LÀ1 h h n i X p ỵ ni ; jT ẳ Q Gn; pịei2Ppj=L NT LT pẳ0 3ị 2 where Gn; pị ¼ eÀ 2P p =n ¼ Gaussian function and where, j; p; n ¼ 0; 1; ; L À The following steps are adapted for computing the discrete S-Transform Perform the discrete Fourier transformÂ of the Ã original time series q½kT (with N p points and sampling interval) to get Q =LT using the FFT routine This is only done once Calculate the localizing Gaussian Gðn; pÞ for the required frequency n=LT h i Â Ã Shift the spectrum Q p=LT to Q p ỵ nị=LT to for the frequency n=LT (one pointer addition) h i h i Multiply Q ðp ỵ nị=LT by Gẵn; p to get A n=LT;p=LT (L multiplications) h i h i Inverse Fourier transform of A n=LT;p=LT p=LT to j to give the row of S n=LT;jT corresponding to the frequency n=LT h i Repeat steps 3, 4, and until all the rows of S n=LT;jT corresponding to all discrete frequencies n=LT; have been deﬁned From (3), it is seen that the output from the S-Transform is an L Â P matrix called the S-matrix whose rows pertain to frequency and columns to time Each element of the S-matrix is complex valued The choice of windowing function is not limited to the Gaussian function; other windowing functions were also implemented successfully The plots of the EEG signals from the ﬁrst electrode and their corresponding S-transforms given in Fig 2.3 Artiﬁcial Neural Networks Artiﬁcial neural networks (ANNs) are computing systems made up of large number of ﬁrmly interconnected adaptive processing elements (neurons) that are able to perform massively parallel computations for data processing and knowledge representation Learning in ANNs is accomplished through special training algorithms developed based on learning rules presumed to mimic the learning mechanisms of Feature Extraction and Performance Analysis of EEG Signal … 669 Fig EEG signals and their corresponding S-transforms biological systems [5, 9] ANNs can be trained to recognize patterns and the nonlinear models developed during training allow neural networks to generalize their conclusions and to make application to patterns not previously encountered Feature Extraction and Classiﬁcation The feature extraction of the above ﬁve sets of data are done by different methods followed by classiﬁcation 3.1 Method-I The features i.e mean, standard deviation, skewness and kurtosis are directly calculated from each electrode data signal of each class of the data sets available Then, some of the data are fed to a neural network for training it to be able to classify those types of EEG signals and the rest data samples are used for testing the accuracy of the classiﬁcation process given Figs 3, (Simulation 1) and (Simulation 2) Table shows simulation result 670 M Swain et al Fig Block diagram of method-1 Fig Error minimization plot [classiﬁcation accuracy (in %): 32] Fig Error minimization plot [classiﬁcation accuracy (in %): 56] 3.2 Method-II The features i.e mean, standard deviation, skewness and kurtosis are calculated after obtaining the S-transform [8] of each electrode data signal of each class of the Total number of electrodes 100 100 Classes used A, B, C, D, E A, E 256 256 200 500 Number of data points Total number used for each electrode of samples Table Simulation and results 100 450 20 20 No of samples No of hidden layer used for training nodes in neural network 70 70 Training percentage 5 Validation percentage 25 25 Testing percentage 0.13241 0.146738 Performance analysis Feature Extraction and Performance Analysis of EEG Signal … 671 672 M Swain et al data sets available Then, some of the data is fed to a neural network for training so as to be able to classify those types of EEG signals and the rest data samples are used for testing the accuracy of the classiﬁcation process given in Figs 6, (Simulation 1) and (Simulation 2) Table shows simulation result Fig Block diagram of method-2 Fig Error minimization plot [classiﬁcation accuracy (in %): 60] Fig Error minimization plot [classiﬁcation accuracy (in %): 94] Total number of electrodes 100 100 Classes used A, B, C, D, E A, E Table Simulation and results 256 256 Number of data points used for each electrode 200 500 Total number of samples 100 450 No of samples used for training 20 20 No of hidden layer nodes in neural network 70 70 Training percentage 5 Validation percentage 25 25 Testing percentage 0.13241 0.146738 Performance analysis Feature Extraction and Performance Analysis of EEG Signal … 673 674 M Swain et al Conclusion Feature extraction is being performed using two methods: S transform and without S transform Performance analysis is evaluated using pattern recognition toolbox In this paper we have extracted features directly from the ﬁve data sets which give a poor percentage of accuracy Hence, we have used S-transform before feature extraction which gives a very good classiﬁcation accuracy score of 94 % References Subasi, A.: EEG signal classiﬁcation using wavelet feature extraction and a mixture of expert model Expert Syst Appl 32, 1084–1093 (2007) Adeli, H., Zhou, Z., Dadmehr, N.: Analysis of EEG records in an epileptic patient using wavelet transform J Neurosci Methods 123, 69–87 (2003) Jahankhani, P., Kodogiannis, V., Revett, K.: EEG signal classiﬁcation using wavelet feature extraction and neural networks International Symposium on Modern Computing, pp 52–57 (2006) Subasi, A.: Automatic recognition of alertness level from EEG by using neural network and wavelet coefﬁcients Expert Syst Appl 28, 701–711 (2005) Subasi, A., Ercelebi, E.: Classiﬁcation of EEG signals using neural network and logistic regression Comput Methods Programs Biomed 78, 87–99 (2005) Petrosian, A., Prokhorov, D., Homan, R., Dashei, R., Wunsch, D.: Recurrent neural network based prediction of epileptic seizures intra and extra cranial EEG Neurocomputing, 30, 201–218 (2000) Jian-feng, H.U.: Multifeature analysis in motor imagery EEG classiﬁcation In: Proceedings of 3rd International Symposium on Electronic Commerce and Security of IEEE, pp 114–117 (2010) Mishra, S., Bhende, C.N., Panigrahi, B.K.: Detection and classiﬁcation of power quality disturbances using S-transform and probabilistic neural network IEEE Trans Power Deliv 23 (1), 280–287 (2008) Krusienski, D.J., McFarland, D.J., Wolpaw, J.R.: An evaluation of autoregressive spectral estimation model order for brain-computer interface applications In: EMBS Annual International Conference of IEEE, New York, pp 1323–1326 (2006) Image Super Resolution Reconstruction Using Iterative Adaptive Regularization Method and Genetic Algorithm S.S Panda, G Jena and S.K Sahu Abstract Super resolution is a technique to obtain high resolution images from several degraded low-resolution images This has got attention in the research society because of its wide use in many ﬁelds of science and technology Even though many methods exist for super resolution, adaptive regularization method is preferred because of its simplicity and the constraints used to get better image restoration result In this paper ﬁrst adaptive algorithm is considered to restore better edge and texture of image Further Genetic algorithm is used to smooth the noise and better frequency addition into the image to get an optimum super resolution image Keywords Peak signal to noise ratio (PSNR) resolution (LR:HR) Genetic algorithm (GA) Á Á Regularization Á Low/high Introduction Image Resolution is an important term in image processing which deals with the quality of various image acquisitions and processing devices Resolution can be deﬁned as the smallest measurable detail in an image In digital image processing we have three different types of image resolution parameters, such as: spatial resolution, brightness resolution and temporal resolution This work considered spatial resolution instead of spectral or temporal resolution for super resolution process, as it says about the spacing of pixels in an image and is measured in pixels per inch (ppi) The higher the spatial resolution of an image, greater the number of pixels in the image accordingly, smaller the size of individual pixels will be The spatial S.S Panda (&) AMET University, Chennai, Tamil Nadu, India e-mail: sudamshekhar@gmail.com G Jena Á S.K Sahu Roland Institute of Technology, Berhampur, Odisha, India e-mail: g_jena@rediffmail.com © Springer India 2015 L.C Jain et al (eds.), Computational Intelligence in Data Mining - Volume 2, Smart Innovation, Systems and Technologies 32, DOI 10.1007/978-81-322-2208-8_62 675 ... Electronics Engineering, Birla Institute of Technology, Mesra, Ranchi, India e-mail: Sudhansu.nit@gmail.com © Springer India 2015 L.C Jain et al (eds.), Computational Intelligence in Data Mining - Volume... http://www.springer.com/series/8767 Lakhmi C Jain Himansu Sekhar Behera Jyotsna Kumar Mandal Durga Prasad Mohapatra • • Editors Computational Intelligence in Data Mining - Volume Proceedings of the International... Intelligence in Data mining for researchers and practitioners • Promote novel high quality research ﬁndings and innovative solutions to the challenging problems in Intelligent Computing • Make
- Xem thêm -
Xem thêm: IT training computational intelligence in data mining (vol 2) jain, behera, mandal mohapatra 2014 12 11 , IT training computational intelligence in data mining (vol 2) jain, behera, mandal mohapatra 2014 12 11 , 2 Expectation Naïve Bayes Classifier---Least Recently Used, 3 Continuous, Discrete, Fast Fourier Transforms